Editorial for the research topic: Digital Twins in Medicine -Transition from Theoretical Concept to Tool used in Everyday Care

Stephen Gilbert ^1* David Drummond ² Fabienne Cotte ³ Tjalf Ziemssen ¹

• ¹ TUD Dresden University of Technology, Dresden, Germany
• ² Service de Pneumologie et d’Allergologie Pédiatriques, Hôpital Necker-Enfants Malades, AP-HP, Paris, France
• ³ Department of Emergency Medicine, University Clinic Marburg, Philipps-University, Marburg, Germany, Marburg, Germany

This Research Topic gathers different contributions addressing the practical advancement of the concept of digital twins in medicine, moving it form a vague theoretical concept towards the foundation to tools used in everyday healthcare. The digital twin (sometimes known as a virtual twin) is a concept that is mainstream in manufacturing, where a digital representation is created of an intended or actual real-world physical product, system, or process (the physical twin). The digital twin serves as an effectively indistinguishable digital counterpart to the physical twin and is used for practical purposes such as simulation, monitoring and maintenance (Singh et al., 2021). This concept has existed in medicine for decades, but unlike in industry, it has not found its way to practical dayto-day application in patient care (Venkatesh et al., 2022;Derraz et al., 2024). Despite this there is renewed research interest in this theme.The goal of this Research Topic was to address if we are at the dawn of the digital twin in medical practice and to explore what is needed to realize this. The articles help to define the aspects of digital twin research that are near to translation and those that need substantially more preclinical development before practical application is possible. Digital Twins of patients, which have been defined in various ways such as "a viewable digital replica of a patient, organ, or biological system that contains multidimensional, patient-specific information and informs decisions" (Drummond and Gonsard, 2024), involve not only new forms of representation of information about patients, but also simulation methods and often AI-based predictive analytical methods. There is much hype and excitation about medical AI, but medical AI will only delivery its promise if firmly linked to the status of the patient in datain other words to the digital twin. These raise regulatory and ethical questions, with differing approaches in differing countries -a goal of the Research Topic was to bring some clarity to these challenges alongside different proposed strategies and developments, and as such to serve as a description of the state of the art and path to impact of digital twins in medicine. This is a provisional file, not the final typesetThe first article of this (Laubenbacher et al., 2024) clinicians and the need for data-driven decision support tools is clear, with some already in use. The 60 authors describe a similarity-based digital twin approach that matches patients with similar historical 61 cases to predict treatment outcomes. Requirements were defined from scientific and technical literature 62 and a four-layer was implemented. The digital twin suggests multi-line treatment strategies with the 63 integration of external evidence with transparency in the data processing logic. The article sets an 64 initial approach to clinical evaluation and illustrates the approach through a detailed description of an 65 exemplary use case for multiple myeloma. 66The third article of this Research Topic (Zhang et al., 2024) is an original article that describes a 67 digital twin framework for type 2 diabetes that integrates machine learning with multiomic data, 68 alongside both knowledge graphs and mechanistic models. The researchers developed predictive 69 machine learning models to forecast disease progression using a substantial dataset comprising clinical 70 measurements and multiomic profiles. Knowledge graphs were employed to interpret and provide 71 context to disease relationships. Promise is demonstrated through the modeling framework reaffirming 72 known targetable disease mechanisms and features. The approach has potential as a DT system for 73 precision medicine. 74The final article of this Research Topic is a mini review of the role for digital patient twins for 75 personalized therapeutics and pharmaceutical manufacturing (Fischer et al., 2024). The authors set out 76 how digital twins of patients pave the way, not only for decision support systems and improved disease 77 monitoring (as described in the previous three articles) but also facilitate the development of 78 personalized therapeutics through their approach to the management, analysis, and interpretation of 79 medical data. The authors identify some gaps that need to be filled before this can be part of routine